diff options
Diffstat (limited to 'vendor/github.com/hashicorp/hcl/hcl')
-rw-r--r-- | vendor/github.com/hashicorp/hcl/hcl/ast/ast.go | 211 | ||||
-rw-r--r-- | vendor/github.com/hashicorp/hcl/hcl/ast/walk.go | 52 | ||||
-rw-r--r-- | vendor/github.com/hashicorp/hcl/hcl/parser/error.go | 17 | ||||
-rw-r--r-- | vendor/github.com/hashicorp/hcl/hcl/parser/parser.go | 454 | ||||
-rw-r--r-- | vendor/github.com/hashicorp/hcl/hcl/scanner/scanner.go | 629 | ||||
-rw-r--r-- | vendor/github.com/hashicorp/hcl/hcl/strconv/quote.go | 245 | ||||
-rw-r--r-- | vendor/github.com/hashicorp/hcl/hcl/token/position.go | 46 | ||||
-rw-r--r-- | vendor/github.com/hashicorp/hcl/hcl/token/token.go | 214 |
8 files changed, 1868 insertions, 0 deletions
diff --git a/vendor/github.com/hashicorp/hcl/hcl/ast/ast.go b/vendor/github.com/hashicorp/hcl/hcl/ast/ast.go new file mode 100644 index 0000000..f8bb71a --- /dev/null +++ b/vendor/github.com/hashicorp/hcl/hcl/ast/ast.go @@ -0,0 +1,211 @@ +// Package ast declares the types used to represent syntax trees for HCL +// (HashiCorp Configuration Language) +package ast + +import ( + "fmt" + "strings" + + "github.com/hashicorp/hcl/hcl/token" +) + +// Node is an element in the abstract syntax tree. +type Node interface { + node() + Pos() token.Pos +} + +func (File) node() {} +func (ObjectList) node() {} +func (ObjectKey) node() {} +func (ObjectItem) node() {} +func (Comment) node() {} +func (CommentGroup) node() {} +func (ObjectType) node() {} +func (LiteralType) node() {} +func (ListType) node() {} + +// File represents a single HCL file +type File struct { + Node Node // usually a *ObjectList + Comments []*CommentGroup // list of all comments in the source +} + +func (f *File) Pos() token.Pos { + return f.Node.Pos() +} + +// ObjectList represents a list of ObjectItems. An HCL file itself is an +// ObjectList. +type ObjectList struct { + Items []*ObjectItem +} + +func (o *ObjectList) Add(item *ObjectItem) { + o.Items = append(o.Items, item) +} + +// Filter filters out the objects with the given key list as a prefix. +// +// The returned list of objects contain ObjectItems where the keys have +// this prefix already stripped off. This might result in objects with +// zero-length key lists if they have no children. +// +// If no matches are found, an empty ObjectList (non-nil) is returned. +func (o *ObjectList) Filter(keys ...string) *ObjectList { + var result ObjectList + for _, item := range o.Items { + // If there aren't enough keys, then ignore this + if len(item.Keys) < len(keys) { + continue + } + + match := true + for i, key := range item.Keys[:len(keys)] { + key := key.Token.Value().(string) + if key != keys[i] && !strings.EqualFold(key, keys[i]) { + match = false + break + } + } + if !match { + continue + } + + // Strip off the prefix from the children + newItem := *item + newItem.Keys = newItem.Keys[len(keys):] + result.Add(&newItem) + } + + return &result +} + +// Children returns further nested objects (key length > 0) within this +// ObjectList. This should be used with Filter to get at child items. +func (o *ObjectList) Children() *ObjectList { + var result ObjectList + for _, item := range o.Items { + if len(item.Keys) > 0 { + result.Add(item) + } + } + + return &result +} + +// Elem returns items in the list that are direct element assignments +// (key length == 0). This should be used with Filter to get at elements. +func (o *ObjectList) Elem() *ObjectList { + var result ObjectList + for _, item := range o.Items { + if len(item.Keys) == 0 { + result.Add(item) + } + } + + return &result +} + +func (o *ObjectList) Pos() token.Pos { + // always returns the uninitiliazed position + return o.Items[0].Pos() +} + +// ObjectItem represents a HCL Object Item. An item is represented with a key +// (or keys). It can be an assignment or an object (both normal and nested) +type ObjectItem struct { + // keys is only one length long if it's of type assignment. If it's a + // nested object it can be larger than one. In that case "assign" is + // invalid as there is no assignments for a nested object. + Keys []*ObjectKey + + // assign contains the position of "=", if any + Assign token.Pos + + // val is the item itself. It can be an object,list, number, bool or a + // string. If key length is larger than one, val can be only of type + // Object. + Val Node + + LeadComment *CommentGroup // associated lead comment + LineComment *CommentGroup // associated line comment +} + +func (o *ObjectItem) Pos() token.Pos { + return o.Keys[0].Pos() +} + +// ObjectKeys are either an identifier or of type string. +type ObjectKey struct { + Token token.Token +} + +func (o *ObjectKey) Pos() token.Pos { + return o.Token.Pos +} + +// LiteralType represents a literal of basic type. Valid types are: +// token.NUMBER, token.FLOAT, token.BOOL and token.STRING +type LiteralType struct { + Token token.Token + + // associated line comment, only when used in a list + LineComment *CommentGroup +} + +func (l *LiteralType) Pos() token.Pos { + return l.Token.Pos +} + +// ListStatement represents a HCL List type +type ListType struct { + Lbrack token.Pos // position of "[" + Rbrack token.Pos // position of "]" + List []Node // the elements in lexical order +} + +func (l *ListType) Pos() token.Pos { + return l.Lbrack +} + +func (l *ListType) Add(node Node) { + l.List = append(l.List, node) +} + +// ObjectType represents a HCL Object Type +type ObjectType struct { + Lbrace token.Pos // position of "{" + Rbrace token.Pos // position of "}" + List *ObjectList // the nodes in lexical order +} + +func (o *ObjectType) Pos() token.Pos { + return o.Lbrace +} + +// Comment node represents a single //, # style or /*- style commment +type Comment struct { + Start token.Pos // position of / or # + Text string +} + +func (c *Comment) Pos() token.Pos { + return c.Start +} + +// CommentGroup node represents a sequence of comments with no other tokens and +// no empty lines between. +type CommentGroup struct { + List []*Comment // len(List) > 0 +} + +func (c *CommentGroup) Pos() token.Pos { + return c.List[0].Pos() +} + +//------------------------------------------------------------------- +// GoStringer +//------------------------------------------------------------------- + +func (o *ObjectKey) GoString() string { return fmt.Sprintf("*%#v", *o) } diff --git a/vendor/github.com/hashicorp/hcl/hcl/ast/walk.go b/vendor/github.com/hashicorp/hcl/hcl/ast/walk.go new file mode 100644 index 0000000..ba07ad4 --- /dev/null +++ b/vendor/github.com/hashicorp/hcl/hcl/ast/walk.go @@ -0,0 +1,52 @@ +package ast + +import "fmt" + +// WalkFunc describes a function to be called for each node during a Walk. The +// returned node can be used to rewrite the AST. Walking stops the returned +// bool is false. +type WalkFunc func(Node) (Node, bool) + +// Walk traverses an AST in depth-first order: It starts by calling fn(node); +// node must not be nil. If fn returns true, Walk invokes fn recursively for +// each of the non-nil children of node, followed by a call of fn(nil). The +// returned node of fn can be used to rewrite the passed node to fn. +func Walk(node Node, fn WalkFunc) Node { + rewritten, ok := fn(node) + if !ok { + return rewritten + } + + switch n := node.(type) { + case *File: + n.Node = Walk(n.Node, fn) + case *ObjectList: + for i, item := range n.Items { + n.Items[i] = Walk(item, fn).(*ObjectItem) + } + case *ObjectKey: + // nothing to do + case *ObjectItem: + for i, k := range n.Keys { + n.Keys[i] = Walk(k, fn).(*ObjectKey) + } + + if n.Val != nil { + n.Val = Walk(n.Val, fn) + } + case *LiteralType: + // nothing to do + case *ListType: + for i, l := range n.List { + n.List[i] = Walk(l, fn) + } + case *ObjectType: + n.List = Walk(n.List, fn).(*ObjectList) + default: + // should we panic here? + fmt.Printf("unknown type: %T\n", n) + } + + fn(nil) + return rewritten +} diff --git a/vendor/github.com/hashicorp/hcl/hcl/parser/error.go b/vendor/github.com/hashicorp/hcl/hcl/parser/error.go new file mode 100644 index 0000000..5c99381 --- /dev/null +++ b/vendor/github.com/hashicorp/hcl/hcl/parser/error.go @@ -0,0 +1,17 @@ +package parser + +import ( + "fmt" + + "github.com/hashicorp/hcl/hcl/token" +) + +// PosError is a parse error that contains a position. +type PosError struct { + Pos token.Pos + Err error +} + +func (e *PosError) Error() string { + return fmt.Sprintf("At %s: %s", e.Pos, e.Err) +} diff --git a/vendor/github.com/hashicorp/hcl/hcl/parser/parser.go b/vendor/github.com/hashicorp/hcl/hcl/parser/parser.go new file mode 100644 index 0000000..cc129b6 --- /dev/null +++ b/vendor/github.com/hashicorp/hcl/hcl/parser/parser.go @@ -0,0 +1,454 @@ +// Package parser implements a parser for HCL (HashiCorp Configuration +// Language) +package parser + +import ( + "errors" + "fmt" + "strings" + + "github.com/hashicorp/hcl/hcl/ast" + "github.com/hashicorp/hcl/hcl/scanner" + "github.com/hashicorp/hcl/hcl/token" +) + +type Parser struct { + sc *scanner.Scanner + + // Last read token + tok token.Token + commaPrev token.Token + + comments []*ast.CommentGroup + leadComment *ast.CommentGroup // last lead comment + lineComment *ast.CommentGroup // last line comment + + enableTrace bool + indent int + n int // buffer size (max = 1) +} + +func newParser(src []byte) *Parser { + return &Parser{ + sc: scanner.New(src), + } +} + +// Parse returns the fully parsed source and returns the abstract syntax tree. +func Parse(src []byte) (*ast.File, error) { + p := newParser(src) + return p.Parse() +} + +var errEofToken = errors.New("EOF token found") + +// Parse returns the fully parsed source and returns the abstract syntax tree. +func (p *Parser) Parse() (*ast.File, error) { + f := &ast.File{} + var err, scerr error + p.sc.Error = func(pos token.Pos, msg string) { + scerr = &PosError{Pos: pos, Err: errors.New(msg)} + } + + f.Node, err = p.objectList() + if scerr != nil { + return nil, scerr + } + if err != nil { + return nil, err + } + + f.Comments = p.comments + return f, nil +} + +func (p *Parser) objectList() (*ast.ObjectList, error) { + defer un(trace(p, "ParseObjectList")) + node := &ast.ObjectList{} + + for { + n, err := p.objectItem() + if err == errEofToken { + break // we are finished + } + + // we don't return a nil node, because might want to use already + // collected items. + if err != nil { + return node, err + } + + node.Add(n) + } + return node, nil +} + +func (p *Parser) consumeComment() (comment *ast.Comment, endline int) { + endline = p.tok.Pos.Line + + // count the endline if it's multiline comment, ie starting with /* + if len(p.tok.Text) > 1 && p.tok.Text[1] == '*' { + // don't use range here - no need to decode Unicode code points + for i := 0; i < len(p.tok.Text); i++ { + if p.tok.Text[i] == '\n' { + endline++ + } + } + } + + comment = &ast.Comment{Start: p.tok.Pos, Text: p.tok.Text} + p.tok = p.sc.Scan() + return +} + +func (p *Parser) consumeCommentGroup(n int) (comments *ast.CommentGroup, endline int) { + var list []*ast.Comment + endline = p.tok.Pos.Line + + for p.tok.Type == token.COMMENT && p.tok.Pos.Line <= endline+n { + var comment *ast.Comment + comment, endline = p.consumeComment() + list = append(list, comment) + } + + // add comment group to the comments list + comments = &ast.CommentGroup{List: list} + p.comments = append(p.comments, comments) + + return +} + +// objectItem parses a single object item +func (p *Parser) objectItem() (*ast.ObjectItem, error) { + defer un(trace(p, "ParseObjectItem")) + + keys, err := p.objectKey() + if len(keys) > 0 && err == errEofToken { + // We ignore eof token here since it is an error if we didn't + // receive a value (but we did receive a key) for the item. + err = nil + } + if len(keys) > 0 && err != nil && p.tok.Type == token.RBRACE { + // This is a strange boolean statement, but what it means is: + // We have keys with no value, and we're likely in an object + // (since RBrace ends an object). For this, we set err to nil so + // we continue and get the error below of having the wrong value + // type. + err = nil + + // Reset the token type so we don't think it completed fine. See + // objectType which uses p.tok.Type to check if we're done with + // the object. + p.tok.Type = token.EOF + } + if err != nil { + return nil, err + } + + o := &ast.ObjectItem{ + Keys: keys, + } + + if p.leadComment != nil { + o.LeadComment = p.leadComment + p.leadComment = nil + } + + switch p.tok.Type { + case token.ASSIGN: + o.Assign = p.tok.Pos + o.Val, err = p.object() + if err != nil { + return nil, err + } + case token.LBRACE: + o.Val, err = p.objectType() + if err != nil { + return nil, err + } + default: + keyStr := make([]string, 0, len(keys)) + for _, k := range keys { + keyStr = append(keyStr, k.Token.Text) + } + + return nil, fmt.Errorf( + "key '%s' expected start of object ('{') or assignment ('=')", + strings.Join(keyStr, " ")) + } + + // do a look-ahead for line comment + p.scan() + if len(keys) > 0 && o.Val.Pos().Line == keys[0].Pos().Line && p.lineComment != nil { + o.LineComment = p.lineComment + p.lineComment = nil + } + p.unscan() + return o, nil +} + +// objectKey parses an object key and returns a ObjectKey AST +func (p *Parser) objectKey() ([]*ast.ObjectKey, error) { + keyCount := 0 + keys := make([]*ast.ObjectKey, 0) + + for { + tok := p.scan() + switch tok.Type { + case token.EOF: + // It is very important to also return the keys here as well as + // the error. This is because we need to be able to tell if we + // did parse keys prior to finding the EOF, or if we just found + // a bare EOF. + return keys, errEofToken + case token.ASSIGN: + // assignment or object only, but not nested objects. this is not + // allowed: `foo bar = {}` + if keyCount > 1 { + return nil, &PosError{ + Pos: p.tok.Pos, + Err: fmt.Errorf("nested object expected: LBRACE got: %s", p.tok.Type), + } + } + + if keyCount == 0 { + return nil, &PosError{ + Pos: p.tok.Pos, + Err: errors.New("no object keys found!"), + } + } + + return keys, nil + case token.LBRACE: + // object + return keys, nil + case token.IDENT, token.STRING: + keyCount++ + keys = append(keys, &ast.ObjectKey{Token: p.tok}) + case token.ILLEGAL: + fmt.Println("illegal") + default: + return keys, &PosError{ + Pos: p.tok.Pos, + Err: fmt.Errorf("expected: IDENT | STRING | ASSIGN | LBRACE got: %s", p.tok.Type), + } + } + } +} + +// object parses any type of object, such as number, bool, string, object or +// list. +func (p *Parser) object() (ast.Node, error) { + defer un(trace(p, "ParseType")) + tok := p.scan() + + switch tok.Type { + case token.NUMBER, token.FLOAT, token.BOOL, token.STRING, token.HEREDOC: + return p.literalType() + case token.LBRACE: + return p.objectType() + case token.LBRACK: + return p.listType() + case token.COMMENT: + // implement comment + case token.EOF: + return nil, errEofToken + } + + return nil, &PosError{ + Pos: tok.Pos, + Err: fmt.Errorf("Unknown token: %+v", tok), + } +} + +// objectType parses an object type and returns a ObjectType AST +func (p *Parser) objectType() (*ast.ObjectType, error) { + defer un(trace(p, "ParseObjectType")) + + // we assume that the currently scanned token is a LBRACE + o := &ast.ObjectType{ + Lbrace: p.tok.Pos, + } + + l, err := p.objectList() + + // if we hit RBRACE, we are good to go (means we parsed all Items), if it's + // not a RBRACE, it's an syntax error and we just return it. + if err != nil && p.tok.Type != token.RBRACE { + return nil, err + } + + // If there is no error, we should be at a RBRACE to end the object + if p.tok.Type != token.RBRACE { + return nil, fmt.Errorf("object expected closing RBRACE got: %s", p.tok.Type) + } + + o.List = l + o.Rbrace = p.tok.Pos // advanced via parseObjectList + return o, nil +} + +// listType parses a list type and returns a ListType AST +func (p *Parser) listType() (*ast.ListType, error) { + defer un(trace(p, "ParseListType")) + + // we assume that the currently scanned token is a LBRACK + l := &ast.ListType{ + Lbrack: p.tok.Pos, + } + + needComma := false + for { + tok := p.scan() + switch tok.Type { + case token.NUMBER, token.FLOAT, token.STRING, token.HEREDOC: + if needComma { + return nil, &PosError{ + Pos: tok.Pos, + Err: fmt.Errorf("unexpected token: %s. Expecting %s", tok.Type, token.COMMA), + } + } + + node, err := p.literalType() + if err != nil { + return nil, err + } + + l.Add(node) + needComma = true + case token.COMMA: + // get next list item or we are at the end + // do a look-ahead for line comment + p.scan() + if p.lineComment != nil { + lit, ok := l.List[len(l.List)-1].(*ast.LiteralType) + if ok { + lit.LineComment = p.lineComment + l.List[len(l.List)-1] = lit + p.lineComment = nil + } + } + p.unscan() + + needComma = false + continue + case token.BOOL: + // TODO(arslan) should we support? not supported by HCL yet + case token.LBRACK: + // TODO(arslan) should we support nested lists? Even though it's + // written in README of HCL, it's not a part of the grammar + // (not defined in parse.y) + case token.RBRACK: + // finished + l.Rbrack = p.tok.Pos + return l, nil + default: + return nil, &PosError{ + Pos: tok.Pos, + Err: fmt.Errorf("unexpected token while parsing list: %s", tok.Type), + } + } + } +} + +// literalType parses a literal type and returns a LiteralType AST +func (p *Parser) literalType() (*ast.LiteralType, error) { + defer un(trace(p, "ParseLiteral")) + + return &ast.LiteralType{ + Token: p.tok, + }, nil +} + +// scan returns the next token from the underlying scanner. If a token has +// been unscanned then read that instead. In the process, it collects any +// comment groups encountered, and remembers the last lead and line comments. +func (p *Parser) scan() token.Token { + // If we have a token on the buffer, then return it. + if p.n != 0 { + p.n = 0 + return p.tok + } + + // Otherwise read the next token from the scanner and Save it to the buffer + // in case we unscan later. + prev := p.tok + p.tok = p.sc.Scan() + + if p.tok.Type == token.COMMENT { + var comment *ast.CommentGroup + var endline int + + // fmt.Printf("p.tok.Pos.Line = %+v prev: %d endline %d \n", + // p.tok.Pos.Line, prev.Pos.Line, endline) + if p.tok.Pos.Line == prev.Pos.Line { + // The comment is on same line as the previous token; it + // cannot be a lead comment but may be a line comment. + comment, endline = p.consumeCommentGroup(0) + if p.tok.Pos.Line != endline { + // The next token is on a different line, thus + // the last comment group is a line comment. + p.lineComment = comment + } + } + + // consume successor comments, if any + endline = -1 + for p.tok.Type == token.COMMENT { + comment, endline = p.consumeCommentGroup(1) + } + + if endline+1 == p.tok.Pos.Line && p.tok.Type != token.RBRACE { + switch p.tok.Type { + case token.RBRACE, token.RBRACK: + // Do not count for these cases + default: + // The next token is following on the line immediately after the + // comment group, thus the last comment group is a lead comment. + p.leadComment = comment + } + } + + } + + return p.tok +} + +// unscan pushes the previously read token back onto the buffer. +func (p *Parser) unscan() { + p.n = 1 +} + +// ---------------------------------------------------------------------------- +// Parsing support + +func (p *Parser) printTrace(a ...interface{}) { + if !p.enableTrace { + return + } + + const dots = ". . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . " + const n = len(dots) + fmt.Printf("%5d:%3d: ", p.tok.Pos.Line, p.tok.Pos.Column) + + i := 2 * p.indent + for i > n { + fmt.Print(dots) + i -= n + } + // i <= n + fmt.Print(dots[0:i]) + fmt.Println(a...) +} + +func trace(p *Parser, msg string) *Parser { + p.printTrace(msg, "(") + p.indent++ + return p +} + +// Usage pattern: defer un(trace(p, "...")) +func un(p *Parser) { + p.indent-- + p.printTrace(")") +} diff --git a/vendor/github.com/hashicorp/hcl/hcl/scanner/scanner.go b/vendor/github.com/hashicorp/hcl/hcl/scanner/scanner.go new file mode 100644 index 0000000..a3f34a7 --- /dev/null +++ b/vendor/github.com/hashicorp/hcl/hcl/scanner/scanner.go @@ -0,0 +1,629 @@ +// Package scanner implements a scanner for HCL (HashiCorp Configuration +// Language) source text. +package scanner + +import ( + "bytes" + "fmt" + "os" + "regexp" + "unicode" + "unicode/utf8" + + "github.com/hashicorp/hcl/hcl/token" +) + +// eof represents a marker rune for the end of the reader. +const eof = rune(0) + +// Scanner defines a lexical scanner +type Scanner struct { + buf *bytes.Buffer // Source buffer for advancing and scanning + src []byte // Source buffer for immutable access + + // Source Position + srcPos token.Pos // current position + prevPos token.Pos // previous position, used for peek() method + + lastCharLen int // length of last character in bytes + lastLineLen int // length of last line in characters (for correct column reporting) + + tokStart int // token text start position + tokEnd int // token text end position + + // Error is called for each error encountered. If no Error + // function is set, the error is reported to os.Stderr. + Error func(pos token.Pos, msg string) + + // ErrorCount is incremented by one for each error encountered. + ErrorCount int + + // tokPos is the start position of most recently scanned token; set by + // Scan. The Filename field is always left untouched by the Scanner. If + // an error is reported (via Error) and Position is invalid, the scanner is + // not inside a token. + tokPos token.Pos +} + +// New creates and initializes a new instance of Scanner using src as +// its source content. +func New(src []byte) *Scanner { + // even though we accept a src, we read from a io.Reader compatible type + // (*bytes.Buffer). So in the future we might easily change it to streaming + // read. + b := bytes.NewBuffer(src) + s := &Scanner{ + buf: b, + src: src, + } + + // srcPosition always starts with 1 + s.srcPos.Line = 1 + return s +} + +// next reads the next rune from the bufferred reader. Returns the rune(0) if +// an error occurs (or io.EOF is returned). +func (s *Scanner) next() rune { + ch, size, err := s.buf.ReadRune() + if err != nil { + // advance for error reporting + s.srcPos.Column++ + s.srcPos.Offset += size + s.lastCharLen = size + return eof + } + + if ch == utf8.RuneError && size == 1 { + s.srcPos.Column++ + s.srcPos.Offset += size + s.lastCharLen = size + s.err("illegal UTF-8 encoding") + return ch + } + + // remember last position + s.prevPos = s.srcPos + + s.srcPos.Column++ + s.lastCharLen = size + s.srcPos.Offset += size + + if ch == '\n' { + s.srcPos.Line++ + s.lastLineLen = s.srcPos.Column + s.srcPos.Column = 0 + } + + // debug + // fmt.Printf("ch: %q, offset:column: %d:%d\n", ch, s.srcPos.Offset, s.srcPos.Column) + return ch +} + +// unread unreads the previous read Rune and updates the source position +func (s *Scanner) unread() { + if err := s.buf.UnreadRune(); err != nil { + panic(err) // this is user fault, we should catch it + } + s.srcPos = s.prevPos // put back last position +} + +// peek returns the next rune without advancing the reader. +func (s *Scanner) peek() rune { + peek, _, err := s.buf.ReadRune() + if err != nil { + return eof + } + + s.buf.UnreadRune() + return peek +} + +// Scan scans the next token and returns the token. +func (s *Scanner) Scan() token.Token { + ch := s.next() + + // skip white space + for isWhitespace(ch) { + ch = s.next() + } + + var tok token.Type + + // token text markings + s.tokStart = s.srcPos.Offset - s.lastCharLen + + // token position, initial next() is moving the offset by one(size of rune + // actually), though we are interested with the starting point + s.tokPos.Offset = s.srcPos.Offset - s.lastCharLen + if s.srcPos.Column > 0 { + // common case: last character was not a '\n' + s.tokPos.Line = s.srcPos.Line + s.tokPos.Column = s.srcPos.Column + } else { + // last character was a '\n' + // (we cannot be at the beginning of the source + // since we have called next() at least once) + s.tokPos.Line = s.srcPos.Line - 1 + s.tokPos.Column = s.lastLineLen + } + + switch { + case isLetter(ch): + tok = token.IDENT + lit := s.scanIdentifier() + if lit == "true" || lit == "false" { + tok = token.BOOL + } + case isDecimal(ch): + tok = s.scanNumber(ch) + default: + switch ch { + case eof: + tok = token.EOF + case '"': + tok = token.STRING + s.scanString() + case '#', '/': + tok = token.COMMENT + s.scanComment(ch) + case '.': + tok = token.PERIOD + ch = s.peek() + if isDecimal(ch) { + tok = token.FLOAT + ch = s.scanMantissa(ch) + ch = s.scanExponent(ch) + } + case '<': + tok = token.HEREDOC + s.scanHeredoc() + case '[': + tok = token.LBRACK + case ']': + tok = token.RBRACK + case '{': + tok = token.LBRACE + case '}': + tok = token.RBRACE + case ',': + tok = token.COMMA + case '=': + tok = token.ASSIGN + case '+': + tok = token.ADD + case '-': + if isDecimal(s.peek()) { + ch := s.next() + tok = s.scanNumber(ch) + } else { + tok = token.SUB + } + default: + s.err("illegal char") + } + } + + // finish token ending + s.tokEnd = s.srcPos.Offset + + // create token literal + var tokenText string + if s.tokStart >= 0 { + tokenText = string(s.src[s.tokStart:s.tokEnd]) + } + s.tokStart = s.tokEnd // ensure idempotency of tokenText() call + + return token.Token{ + Type: tok, + Pos: s.tokPos, + Text: tokenText, + } +} + +func (s *Scanner) scanComment(ch rune) { + // single line comments + if ch == '#' || (ch == '/' && s.peek() != '*') { + ch = s.next() + for ch != '\n' && ch >= 0 && ch != eof { + ch = s.next() + } + if ch != eof && ch >= 0 { + s.unread() + } + return + } + + // be sure we get the character after /* This allows us to find comment's + // that are not erminated + if ch == '/' { + s.next() + ch = s.next() // read character after "/*" + } + + // look for /* - style comments + for { + if ch < 0 || ch == eof { + s.err("comment not terminated") + break + } + + ch0 := ch + ch = s.next() + if ch0 == '*' && ch == '/' { + break + } + } +} + +// scanNumber scans a HCL number definition starting with the given rune +func (s *Scanner) scanNumber(ch rune) token.Type { + if ch == '0' { + // check for hexadecimal, octal or float + ch = s.next() + if ch == 'x' || ch == 'X' { + // hexadecimal + ch = s.next() + found := false + for isHexadecimal(ch) { + ch = s.next() + found = true + } + + if !found { + s.err("illegal hexadecimal number") + } + + if ch != eof { + s.unread() + } + + return token.NUMBER + } + + // now it's either something like: 0421(octal) or 0.1231(float) + illegalOctal := false + for isDecimal(ch) { + ch = s.next() + if ch == '8' || ch == '9' { + // this is just a possibility. For example 0159 is illegal, but + // 0159.23 is valid. So we mark a possible illegal octal. If + // the next character is not a period, we'll print the error. + illegalOctal = true + } + } + + if ch == 'e' || ch == 'E' { + ch = s.scanExponent(ch) + return token.FLOAT + } + + if ch == '.' { + ch = s.scanFraction(ch) + + if ch == 'e' || ch == 'E' { + ch = s.next() + ch = s.scanExponent(ch) + } + return token.FLOAT + } + + if illegalOctal { + s.err("illegal octal number") + } + + if ch != eof { + s.unread() + } + return token.NUMBER + } + + s.scanMantissa(ch) + ch = s.next() // seek forward + if ch == 'e' || ch == 'E' { + ch = s.scanExponent(ch) + return token.FLOAT + } + + if ch == '.' { + ch = s.scanFraction(ch) + if ch == 'e' || ch == 'E' { + ch = s.next() + ch = s.scanExponent(ch) + } + return token.FLOAT + } + + if ch != eof { + s.unread() + } + return token.NUMBER +} + +// scanMantissa scans the mantissa begining from the rune. It returns the next +// non decimal rune. It's used to determine wheter it's a fraction or exponent. +func (s *Scanner) scanMantissa(ch rune) rune { + scanned := false + for isDecimal(ch) { + ch = s.next() + scanned = true + } + + if scanned && ch != eof { + s.unread() + } + return ch +} + +// scanFraction scans the fraction after the '.' rune +func (s *Scanner) scanFraction(ch rune) rune { + if ch == '.' { + ch = s.peek() // we peek just to see if we can move forward + ch = s.scanMantissa(ch) + } + return ch +} + +// scanExponent scans the remaining parts of an exponent after the 'e' or 'E' +// rune. +func (s *Scanner) scanExponent(ch rune) rune { + if ch == 'e' || ch == 'E' { + ch = s.next() + if ch == '-' || ch == '+' { + ch = s.next() + } + ch = s.scanMantissa(ch) + } + return ch +} + +// scanHeredoc scans a heredoc string +func (s *Scanner) scanHeredoc() { + // Scan the second '<' in example: '<<EOF' + if s.next() != '<' { + s.err("heredoc expected second '<', didn't see it") + return + } + + // Get the original offset so we can read just the heredoc ident + offs := s.srcPos.Offset + + // Scan the identifier + ch := s.next() + + // Indented heredoc syntax + if ch == '-' { + ch = s.next() + } + + for isLetter(ch) || isDigit(ch) { + ch = s.next() + } + + // If we reached an EOF then that is not good + if ch == eof { + s.err("heredoc not terminated") + return + } + + // Ignore the '\r' in Windows line endings + if ch == '\r' { + if s.peek() == '\n' { + ch = s.next() + } + } + + // If we didn't reach a newline then that is also not good + if ch != '\n' { + s.err("invalid characters in heredoc anchor") + return + } + + // Read the identifier + identBytes := s.src[offs : s.srcPos.Offset-s.lastCharLen] + if len(identBytes) == 0 { + s.err("zero-length heredoc anchor") + return + } + + var identRegexp *regexp.Regexp + if identBytes[0] == '-' { + identRegexp = regexp.MustCompile(fmt.Sprintf(`[[:space:]]*%s\z`, identBytes[1:])) + } else { + identRegexp = regexp.MustCompile(fmt.Sprintf(`[[:space:]]*%s\z`, identBytes)) + } + + // Read the actual string value + lineStart := s.srcPos.Offset + for { + ch := s.next() + + // Special newline handling. + if ch == '\n' { + // Math is fast, so we first compare the byte counts to see if we have a chance + // of seeing the same identifier - if the length is less than the number of bytes + // in the identifier, this cannot be a valid terminator. + lineBytesLen := s.srcPos.Offset - s.lastCharLen - lineStart + if lineBytesLen >= len(identBytes) && identRegexp.Match(s.src[lineStart:s.srcPos.Offset-s.lastCharLen]) { + break + } + + // Not an anchor match, record the start of a new line + lineStart = s.srcPos.Offset + } + + if ch == eof { + s.err("heredoc not terminated") + return + } + } + + return +} + +// scanString scans a quoted string +func (s *Scanner) scanString() { + braces := 0 + for { + // '"' opening already consumed + // read character after quote + ch := s.next() + + if ch == '\n' || ch < 0 || ch == eof { + s.err("literal not terminated") + return + } + + if ch == '"' && braces == 0 { + break + } + + // If we're going into a ${} then we can ignore quotes for awhile + if braces == 0 && ch == '$' && s.peek() == '{' { + braces++ + s.next() + } else if braces > 0 && ch == '{' { + braces++ + } + if braces > 0 && ch == '}' { + braces-- + } + + if ch == '\\' { + s.scanEscape() + } + } + + return +} + +// scanEscape scans an escape sequence +func (s *Scanner) scanEscape() rune { + // http://en.cppreference.com/w/cpp/language/escape + ch := s.next() // read character after '/' + switch ch { + case 'a', 'b', 'f', 'n', 'r', 't', 'v', '\\', '"': + // nothing to do + case '0', '1', '2', '3', '4', '5', '6', '7': + // octal notation + ch = s.scanDigits(ch, 8, 3) + case 'x': + // hexademical notation + ch = s.scanDigits(s.next(), 16, 2) + case 'u': + // universal character name + ch = s.scanDigits(s.next(), 16, 4) + case 'U': + // universal character name + ch = s.scanDigits(s.next(), 16, 8) + default: + s.err("illegal char escape") + } + return ch +} + +// scanDigits scans a rune with the given base for n times. For example an +// octal notation \184 would yield in scanDigits(ch, 8, 3) +func (s *Scanner) scanDigits(ch rune, base, n int) rune { + for n > 0 && digitVal(ch) < base { + ch = s.next() + n-- + } + if n > 0 { + s.err("illegal char escape") + } + + // we scanned all digits, put the last non digit char back + s.unread() + return ch +} + +// scanIdentifier scans an identifier and returns the literal string +func (s *Scanner) scanIdentifier() string { + offs := s.srcPos.Offset - s.lastCharLen + ch := s.next() + for isLetter(ch) || isDigit(ch) || ch == '-' || ch == '.' { + ch = s.next() + } + + if ch != eof { + s.unread() // we got identifier, put back latest char + } + + return string(s.src[offs:s.srcPos.Offset]) +} + +// recentPosition returns the position of the character immediately after the +// character or token returned by the last call to Scan. +func (s *Scanner) recentPosition() (pos token.Pos) { + pos.Offset = s.srcPos.Offset - s.lastCharLen + switch { + case s.srcPos.Column > 0: + // common case: last character was not a '\n' + pos.Line = s.srcPos.Line + pos.Column = s.srcPos.Column + case s.lastLineLen > 0: + // last character was a '\n' + // (we cannot be at the beginning of the source + // since we have called next() at least once) + pos.Line = s.srcPos.Line - 1 + pos.Column = s.lastLineLen + default: + // at the beginning of the source + pos.Line = 1 + pos.Column = 1 + } + return +} + +// err prints the error of any scanning to s.Error function. If the function is +// not defined, by default it prints them to os.Stderr +func (s *Scanner) err(msg string) { + s.ErrorCount++ + pos := s.recentPosition() + + if s.Error != nil { + s.Error(pos, msg) + return + } + + fmt.Fprintf(os.Stderr, "%s: %s\n", pos, msg) +} + +// isHexadecimal returns true if the given rune is a letter +func isLetter(ch rune) bool { + return 'a' <= ch && ch <= 'z' || 'A' <= ch && ch <= 'Z' || ch == '_' || ch >= 0x80 && unicode.IsLetter(ch) +} + +// isDigit returns true if the given rune is a decimal digit +func isDigit(ch rune) bool { + return '0' <= ch && ch <= '9' || ch >= 0x80 && unicode.IsDigit(ch) +} + +// isDecimal returns true if the given rune is a decimal number +func isDecimal(ch rune) bool { + return '0' <= ch && ch <= '9' +} + +// isHexadecimal returns true if the given rune is an hexadecimal number +func isHexadecimal(ch rune) bool { + return '0' <= ch && ch <= '9' || 'a' <= ch && ch <= 'f' || 'A' <= ch && ch <= 'F' +} + +// isWhitespace returns true if the rune is a space, tab, newline or carriage return +func isWhitespace(ch rune) bool { + return ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r' +} + +// digitVal returns the integer value of a given octal,decimal or hexadecimal rune +func digitVal(ch rune) int { + switch { + case '0' <= ch && ch <= '9': + return int(ch - '0') + case 'a' <= ch && ch <= 'f': + return int(ch - 'a' + 10) + case 'A' <= ch && ch <= 'F': + return int(ch - 'A' + 10) + } + return 16 // larger than any legal digit val +} diff --git a/vendor/github.com/hashicorp/hcl/hcl/strconv/quote.go b/vendor/github.com/hashicorp/hcl/hcl/strconv/quote.go new file mode 100644 index 0000000..e87ac63 --- /dev/null +++ b/vendor/github.com/hashicorp/hcl/hcl/strconv/quote.go @@ -0,0 +1,245 @@ +package strconv + +import ( + "errors" + "unicode/utf8" +) + +// ErrSyntax indicates that a value does not have the right syntax for the target type. +var ErrSyntax = errors.New("invalid syntax") + +// Unquote interprets s as a single-quoted, double-quoted, +// or backquoted Go string literal, returning the string value +// that s quotes. (If s is single-quoted, it would be a Go +// character literal; Unquote returns the corresponding +// one-character string.) +func Unquote(s string) (t string, err error) { + n := len(s) + if n < 2 { + return "", ErrSyntax + } + quote := s[0] + if quote != s[n-1] { + return "", ErrSyntax + } + s = s[1 : n-1] + + if quote != '"' { + return "", ErrSyntax + } + if contains(s, '\n') { + return "", ErrSyntax + } + + // Is it trivial? Avoid allocation. + if !contains(s, '\\') && !contains(s, quote) && !contains(s, '$') { + switch quote { + case '"': + return s, nil + case '\'': + r, size := utf8.DecodeRuneInString(s) + if size == len(s) && (r != utf8.RuneError || size != 1) { + return s, nil + } + } + } + + var runeTmp [utf8.UTFMax]byte + buf := make([]byte, 0, 3*len(s)/2) // Try to avoid more allocations. + for len(s) > 0 { + // If we're starting a '${}' then let it through un-unquoted. + // Specifically: we don't unquote any characters within the `${}` + // section, except for escaped quotes, which we handle specifically. + if s[0] == '$' && len(s) > 1 && s[1] == '{' { + buf = append(buf, '$', '{') + s = s[2:] + + // Continue reading until we find the closing brace, copying as-is + braces := 1 + for len(s) > 0 && braces > 0 { + r, size := utf8.DecodeRuneInString(s) + if r == utf8.RuneError { + return "", ErrSyntax + } + + s = s[size:] + + // We special case escaped double quotes in interpolations, converting + // them to straight double quotes. + if r == '\\' { + if q, _ := utf8.DecodeRuneInString(s); q == '"' { + continue + } + } + + n := utf8.EncodeRune(runeTmp[:], r) + buf = append(buf, runeTmp[:n]...) + + switch r { + case '{': + braces++ + case '}': + braces-- + } + } + if braces != 0 { + return "", ErrSyntax + } + if len(s) == 0 { + // If there's no string left, we're done! + break + } else { + // If there's more left, we need to pop back up to the top of the loop + // in case there's another interpolation in this string. + continue + } + } + + c, multibyte, ss, err := unquoteChar(s, quote) + if err != nil { + return "", err + } + s = ss + if c < utf8.RuneSelf || !multibyte { + buf = append(buf, byte(c)) + } else { + n := utf8.EncodeRune(runeTmp[:], c) + buf = append(buf, runeTmp[:n]...) + } + if quote == '\'' && len(s) != 0 { + // single-quoted must be single character + return "", ErrSyntax + } + } + return string(buf), nil +} + +// contains reports whether the string contains the byte c. +func contains(s string, c byte) bool { + for i := 0; i < len(s); i++ { + if s[i] == c { + return true + } + } + return false +} + +func unhex(b byte) (v rune, ok bool) { + c := rune(b) + switch { + case '0' <= c && c <= '9': + return c - '0', true + case 'a' <= c && c <= 'f': + return c - 'a' + 10, true + case 'A' <= c && c <= 'F': + return c - 'A' + 10, true + } + return +} + +func unquoteChar(s string, quote byte) (value rune, multibyte bool, tail string, err error) { + // easy cases + switch c := s[0]; { + case c == quote && (quote == '\'' || quote == '"'): + err = ErrSyntax + return + case c >= utf8.RuneSelf: + r, size := utf8.DecodeRuneInString(s) + return r, true, s[size:], nil + case c != '\\': + return rune(s[0]), false, s[1:], nil + } + + // hard case: c is backslash + if len(s) <= 1 { + err = ErrSyntax + return + } + c := s[1] + s = s[2:] + + switch c { + case 'a': + value = '\a' + case 'b': + value = '\b' + case 'f': + value = '\f' + case 'n': + value = '\n' + case 'r': + value = '\r' + case 't': + value = '\t' + case 'v': + value = '\v' + case 'x', 'u', 'U': + n := 0 + switch c { + case 'x': + n = 2 + case 'u': + n = 4 + case 'U': + n = 8 + } + var v rune + if len(s) < n { + err = ErrSyntax + return + } + for j := 0; j < n; j++ { + x, ok := unhex(s[j]) + if !ok { + err = ErrSyntax + return + } + v = v<<4 | x + } + s = s[n:] + if c == 'x' { + // single-byte string, possibly not UTF-8 + value = v + break + } + if v > utf8.MaxRune { + err = ErrSyntax + return + } + value = v + multibyte = true + case '0', '1', '2', '3', '4', '5', '6', '7': + v := rune(c) - '0' + if len(s) < 2 { + err = ErrSyntax + return + } + for j := 0; j < 2; j++ { // one digit already; two more + x := rune(s[j]) - '0' + if x < 0 || x > 7 { + err = ErrSyntax + return + } + v = (v << 3) | x + } + s = s[2:] + if v > 255 { + err = ErrSyntax + return + } + value = v + case '\\': + value = '\\' + case '\'', '"': + if c != quote { + err = ErrSyntax + return + } + value = rune(c) + default: + err = ErrSyntax + return + } + tail = s + return +} diff --git a/vendor/github.com/hashicorp/hcl/hcl/token/position.go b/vendor/github.com/hashicorp/hcl/hcl/token/position.go new file mode 100644 index 0000000..59c1bb7 --- /dev/null +++ b/vendor/github.com/hashicorp/hcl/hcl/token/position.go @@ -0,0 +1,46 @@ +package token + +import "fmt" + +// Pos describes an arbitrary source position +// including the file, line, and column location. +// A Position is valid if the line number is > 0. +type Pos struct { + Filename string // filename, if any + Offset int // offset, starting at 0 + Line int // line number, starting at 1 + Column int // column number, starting at 1 (character count) +} + +// IsValid returns true if the position is valid. +func (p *Pos) IsValid() bool { return p.Line > 0 } + +// String returns a string in one of several forms: +// +// file:line:column valid position with file name +// line:column valid position without file name +// file invalid position with file name +// - invalid position without file name +func (p Pos) String() string { + s := p.Filename + if p.IsValid() { + if s != "" { + s += ":" + } + s += fmt.Sprintf("%d:%d", p.Line, p.Column) + } + if s == "" { + s = "-" + } + return s +} + +// Before reports whether the position p is before u. +func (p Pos) Before(u Pos) bool { + return u.Offset > p.Offset || u.Line > p.Line +} + +// After reports whether the position p is after u. +func (p Pos) After(u Pos) bool { + return u.Offset < p.Offset || u.Line < p.Line +} diff --git a/vendor/github.com/hashicorp/hcl/hcl/token/token.go b/vendor/github.com/hashicorp/hcl/hcl/token/token.go new file mode 100644 index 0000000..6e99498 --- /dev/null +++ b/vendor/github.com/hashicorp/hcl/hcl/token/token.go @@ -0,0 +1,214 @@ +// Package token defines constants representing the lexical tokens for HCL +// (HashiCorp Configuration Language) +package token + +import ( + "fmt" + "strconv" + "strings" + + hclstrconv "github.com/hashicorp/hcl/hcl/strconv" +) + +// Token defines a single HCL token which can be obtained via the Scanner +type Token struct { + Type Type + Pos Pos + Text string + JSON bool +} + +// Type is the set of lexical tokens of the HCL (HashiCorp Configuration Language) +type Type int + +const ( + // Special tokens + ILLEGAL Type = iota + EOF + COMMENT + + identifier_beg + IDENT // literals + literal_beg + NUMBER // 12345 + FLOAT // 123.45 + BOOL // true,false + STRING // "abc" + HEREDOC // <<FOO\nbar\nFOO + literal_end + identifier_end + + operator_beg + LBRACK // [ + LBRACE // { + COMMA // , + PERIOD // . + + RBRACK // ] + RBRACE // } + + ASSIGN // = + ADD // + + SUB // - + operator_end +) + +var tokens = [...]string{ + ILLEGAL: "ILLEGAL", + + EOF: "EOF", + COMMENT: "COMMENT", + + IDENT: "IDENT", + NUMBER: "NUMBER", + FLOAT: "FLOAT", + BOOL: "BOOL", + STRING: "STRING", + + LBRACK: "LBRACK", + LBRACE: "LBRACE", + COMMA: "COMMA", + PERIOD: "PERIOD", + HEREDOC: "HEREDOC", + + RBRACK: "RBRACK", + RBRACE: "RBRACE", + + ASSIGN: "ASSIGN", + ADD: "ADD", + SUB: "SUB", +} + +// String returns the string corresponding to the token tok. +func (t Type) String() string { + s := "" + if 0 <= t && t < Type(len(tokens)) { + s = tokens[t] + } + if s == "" { + s = "token(" + strconv.Itoa(int(t)) + ")" + } + return s +} + +// IsIdentifier returns true for tokens corresponding to identifiers and basic +// type literals; it returns false otherwise. +func (t Type) IsIdentifier() bool { return identifier_beg < t && t < identifier_end } + +// IsLiteral returns true for tokens corresponding to basic type literals; it +// returns false otherwise. +func (t Type) IsLiteral() bool { return literal_beg < t && t < literal_end } + +// IsOperator returns true for tokens corresponding to operators and +// delimiters; it returns false otherwise. +func (t Type) IsOperator() bool { return operator_beg < t && t < operator_end } + +// String returns the token's literal text. Note that this is only +// applicable for certain token types, such as token.IDENT, +// token.STRING, etc.. +func (t Token) String() string { + return fmt.Sprintf("%s %s %s", t.Pos.String(), t.Type.String(), t.Text) +} + +// Value returns the properly typed value for this token. The type of +// the returned interface{} is guaranteed based on the Type field. +// +// This can only be called for literal types. If it is called for any other +// type, this will panic. +func (t Token) Value() interface{} { + switch t.Type { + case BOOL: + if t.Text == "true" { + return true + } else if t.Text == "false" { + return false + } + + panic("unknown bool value: " + t.Text) + case FLOAT: + v, err := strconv.ParseFloat(t.Text, 64) + if err != nil { + panic(err) + } + + return float64(v) + case NUMBER: + v, err := strconv.ParseInt(t.Text, 0, 64) + if err != nil { + panic(err) + } + + return int64(v) + case IDENT: + return t.Text + case HEREDOC: + return unindentHeredoc(t.Text) + case STRING: + // Determine the Unquote method to use. If it came from JSON, + // then we need to use the built-in unquote since we have to + // escape interpolations there. + f := hclstrconv.Unquote + if t.JSON { + f = strconv.Unquote + } + + v, err := f(t.Text) + if err != nil { + panic(fmt.Sprintf("unquote %s err: %s", t.Text, err)) + } + + return v + default: + panic(fmt.Sprintf("unimplemented Value for type: %s", t.Type)) + } +} + +// unindentHeredoc returns the string content of a HEREDOC if it is started with << +// and the content of a HEREDOC with the hanging indent removed if it is started with +// a <<-, and the terminating line is at least as indented as the least indented line. +func unindentHeredoc(heredoc string) string { + // We need to find the end of the marker + idx := strings.IndexByte(heredoc, '\n') + if idx == -1 { + panic("heredoc doesn't contain newline") + } + + unindent := heredoc[2] == '-' + + // We can optimize if the heredoc isn't marked for indentation + if !unindent { + return string(heredoc[idx+1 : len(heredoc)-idx+1]) + } + + // We need to unindent each line based on the indentation level of the marker + lines := strings.Split(string(heredoc[idx+1:len(heredoc)-idx+2]), "\n") + whitespacePrefix := lines[len(lines)-1] + + isIndented := true + for _, v := range lines { + if strings.HasPrefix(v, whitespacePrefix) { + continue + } + + isIndented = false + break + } + + // If all lines are not at least as indented as the terminating mark, return the + // heredoc as is, but trim the leading space from the marker on the final line. + if !isIndented { + return strings.TrimRight(string(heredoc[idx+1:len(heredoc)-idx+1]), " \t") + } + + unindentedLines := make([]string, len(lines)) + for k, v := range lines { + if k == len(lines)-1 { + unindentedLines[k] = "" + break + } + + unindentedLines[k] = strings.TrimPrefix(v, whitespacePrefix) + } + + return strings.Join(unindentedLines, "\n") +} |